1 #ifndef _LINUX_MMZONE_H
2 #define _LINUX_MMZONE_H
5 #ifndef __GENERATING_BOUNDS_H
7 #include <linux/spinlock.h>
8 #include <linux/list.h>
9 #include <linux/wait.h>
10 #include <linux/bitops.h>
11 #include <linux/cache.h>
12 #include <linux/threads.h>
13 #include <linux/numa.h>
14 #include <linux/init.h>
15 #include <linux/seqlock.h>
16 #include <linux/nodemask.h>
17 #include <linux/pageblock-flags.h>
18 #include <generated/bounds.h>
19 #include <linux/atomic.h>
22 /* Free memory management - zoned buddy allocator. */
23 #ifndef CONFIG_FORCE_MAX_ZONEORDER
26 #define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER
28 #define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1))
31 * PAGE_ALLOC_COSTLY_ORDER is the order at which allocations are deemed
32 * costly to service. That is between allocation orders which should
33 * coelesce naturally under reasonable reclaim pressure and those which
36 #define PAGE_ALLOC_COSTLY_ORDER 3
38 #define MIGRATE_UNMOVABLE 0
39 #define MIGRATE_RECLAIMABLE 1
40 #define MIGRATE_MOVABLE 2
41 #define MIGRATE_PCPTYPES 3 /* the number of types on the pcp lists */
42 #define MIGRATE_RESERVE 3
43 #define MIGRATE_ISOLATE 4 /* can't allocate from here */
44 #define MIGRATE_TYPES 5
46 #define for_each_migratetype_order(order, type) \
47 for (order = 0; order < MAX_ORDER; order++) \
48 for (type = 0; type < MIGRATE_TYPES; type++)
50 extern int page_group_by_mobility_disabled;
52 static inline int get_pageblock_migratetype(struct page *page)
54 return get_pageblock_flags_group(page, PB_migrate, PB_migrate_end);
58 struct list_head free_list[MIGRATE_TYPES];
59 unsigned long nr_free;
65 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
66 * So add a wild amount of padding here to ensure that they fall into separate
67 * cachelines. There are very few zone structures in the machine, so space
68 * consumption is not a concern here.
70 #if defined(CONFIG_SMP)
73 } ____cacheline_internodealigned_in_smp;
74 #define ZONE_PADDING(name) struct zone_padding name;
76 #define ZONE_PADDING(name)
80 /* First 128 byte cacheline (assuming 64 bit words) */
83 NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */
84 NR_ACTIVE_ANON, /* " " " " " */
85 NR_INACTIVE_FILE, /* " " " " " */
86 NR_ACTIVE_FILE, /* " " " " " */
87 NR_UNEVICTABLE, /* " " " " " */
88 NR_MLOCK, /* mlock()ed pages found and moved off LRU */
89 NR_ANON_PAGES, /* Mapped anonymous pages */
90 NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
91 only modified from process context */
96 NR_SLAB_UNRECLAIMABLE,
97 NR_PAGETABLE, /* used for pagetables */
99 /* Second 128 byte cacheline */
100 NR_UNSTABLE_NFS, /* NFS unstable pages */
103 NR_VMSCAN_IMMEDIATE, /* Prioritise for reclaim when writeback ends */
104 NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */
105 NR_ISOLATED_ANON, /* Temporary isolated pages from anon lru */
106 NR_ISOLATED_FILE, /* Temporary isolated pages from file lru */
107 NR_SHMEM, /* shmem pages (included tmpfs/GEM pages) */
108 NR_DIRTIED, /* page dirtyings since bootup */
109 NR_WRITTEN, /* page writings since bootup */
111 NUMA_HIT, /* allocated in intended node */
112 NUMA_MISS, /* allocated in non intended node */
113 NUMA_FOREIGN, /* was intended here, hit elsewhere */
114 NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */
115 NUMA_LOCAL, /* allocation from local node */
116 NUMA_OTHER, /* allocation from other node */
118 NR_ANON_TRANSPARENT_HUGEPAGES,
119 NR_VM_ZONE_STAT_ITEMS };
122 * We do arithmetic on the LRU lists in various places in the code,
123 * so it is important to keep the active lists LRU_ACTIVE higher in
124 * the array than the corresponding inactive lists, and to keep
125 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
127 * This has to be kept in sync with the statistics in zone_stat_item
128 * above and the descriptions in vmstat_text in mm/vmstat.c
135 LRU_INACTIVE_ANON = LRU_BASE,
136 LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE,
137 LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE,
138 LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE,
143 #define for_each_lru(l) for (l = 0; l < NR_LRU_LISTS; l++)
145 #define for_each_evictable_lru(l) for (l = 0; l <= LRU_ACTIVE_FILE; l++)
147 static inline int is_file_lru(enum lru_list l)
149 return (l == LRU_INACTIVE_FILE || l == LRU_ACTIVE_FILE);
152 static inline int is_active_lru(enum lru_list l)
154 return (l == LRU_ACTIVE_ANON || l == LRU_ACTIVE_FILE);
157 static inline int is_unevictable_lru(enum lru_list l)
159 return (l == LRU_UNEVICTABLE);
162 /* Mask used at gathering information at once (see memcontrol.c) */
163 #define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE))
164 #define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON))
165 #define LRU_ALL_EVICTABLE (LRU_ALL_FILE | LRU_ALL_ANON)
166 #define LRU_ALL ((1 << NR_LRU_LISTS) - 1)
168 /* Isolate inactive pages */
169 #define ISOLATE_INACTIVE ((__force isolate_mode_t)0x1)
170 /* Isolate active pages */
171 #define ISOLATE_ACTIVE ((__force isolate_mode_t)0x2)
172 /* Isolate clean file */
173 #define ISOLATE_CLEAN ((__force isolate_mode_t)0x4)
174 /* Isolate unmapped file */
175 #define ISOLATE_UNMAPPED ((__force isolate_mode_t)0x8)
177 /* LRU Isolation modes. */
178 typedef unsigned __bitwise__ isolate_mode_t;
180 enum zone_watermarks {
187 #define min_wmark_pages(z) (z->watermark[WMARK_MIN])
188 #define low_wmark_pages(z) (z->watermark[WMARK_LOW])
189 #define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
191 struct per_cpu_pages {
192 int count; /* number of pages in the list */
193 int high; /* high watermark, emptying needed */
194 int batch; /* chunk size for buddy add/remove */
196 /* Lists of pages, one per migrate type stored on the pcp-lists */
197 struct list_head lists[MIGRATE_PCPTYPES];
200 struct per_cpu_pageset {
201 struct per_cpu_pages pcp;
207 s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
211 #endif /* !__GENERATING_BOUNDS.H */
214 #ifdef CONFIG_ZONE_DMA
216 * ZONE_DMA is used when there are devices that are not able
217 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
218 * carve out the portion of memory that is needed for these devices.
219 * The range is arch specific.
224 * ---------------------------
225 * parisc, ia64, sparc <4G
228 * alpha Unlimited or 0-16MB.
230 * i386, x86_64 and multiple other arches
235 #ifdef CONFIG_ZONE_DMA32
237 * x86_64 needs two ZONE_DMAs because it supports devices that are
238 * only able to do DMA to the lower 16M but also 32 bit devices that
239 * can only do DMA areas below 4G.
244 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
245 * performed on pages in ZONE_NORMAL if the DMA devices support
246 * transfers to all addressable memory.
249 #ifdef CONFIG_HIGHMEM
251 * A memory area that is only addressable by the kernel through
252 * mapping portions into its own address space. This is for example
253 * used by i386 to allow the kernel to address the memory beyond
254 * 900MB. The kernel will set up special mappings (page
255 * table entries on i386) for each page that the kernel needs to
264 #ifndef __GENERATING_BOUNDS_H
267 * When a memory allocation must conform to specific limitations (such
268 * as being suitable for DMA) the caller will pass in hints to the
269 * allocator in the gfp_mask, in the zone modifier bits. These bits
270 * are used to select a priority ordered list of memory zones which
271 * match the requested limits. See gfp_zone() in include/linux/gfp.h
275 #define ZONES_SHIFT 0
276 #elif MAX_NR_ZONES <= 2
277 #define ZONES_SHIFT 1
278 #elif MAX_NR_ZONES <= 4
279 #define ZONES_SHIFT 2
281 #error ZONES_SHIFT -- too many zones configured adjust calculation
284 struct zone_reclaim_stat {
286 * The pageout code in vmscan.c keeps track of how many of the
287 * mem/swap backed and file backed pages are refeferenced.
288 * The higher the rotated/scanned ratio, the more valuable
291 * The anon LRU stats live in [0], file LRU stats in [1]
293 unsigned long recent_rotated[2];
294 unsigned long recent_scanned[2];
298 /* Fields commonly accessed by the page allocator */
300 /* zone watermarks, access with *_wmark_pages(zone) macros */
301 unsigned long watermark[NR_WMARK];
304 * When free pages are below this point, additional steps are taken
305 * when reading the number of free pages to avoid per-cpu counter
306 * drift allowing watermarks to be breached
308 unsigned long percpu_drift_mark;
311 * We don't know if the memory that we're going to allocate will be freeable
312 * or/and it will be released eventually, so to avoid totally wasting several
313 * GB of ram we must reserve some of the lower zone memory (otherwise we risk
314 * to run OOM on the lower zones despite there's tons of freeable ram
315 * on the higher zones). This array is recalculated at runtime if the
316 * sysctl_lowmem_reserve_ratio sysctl changes.
318 unsigned long lowmem_reserve[MAX_NR_ZONES];
321 * This is a per-zone reserve of pages that should not be
322 * considered dirtyable memory.
324 unsigned long dirty_balance_reserve;
329 * zone reclaim becomes active if more unmapped pages exist.
331 unsigned long min_unmapped_pages;
332 unsigned long min_slab_pages;
334 struct per_cpu_pageset __percpu *pageset;
336 * free areas of different sizes
339 int all_unreclaimable; /* All pages pinned */
340 #ifdef CONFIG_MEMORY_HOTPLUG
341 /* see spanned/present_pages for more description */
342 seqlock_t span_seqlock;
344 struct free_area free_area[MAX_ORDER];
346 #ifndef CONFIG_SPARSEMEM
348 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
349 * In SPARSEMEM, this map is stored in struct mem_section
351 unsigned long *pageblock_flags;
352 #endif /* CONFIG_SPARSEMEM */
354 #ifdef CONFIG_COMPACTION
356 * On compaction failure, 1<<compact_defer_shift compactions
357 * are skipped before trying again. The number attempted since
358 * last failure is tracked with compact_considered.
360 unsigned int compact_considered;
361 unsigned int compact_defer_shift;
366 /* Fields commonly accessed by the page reclaim scanner */
369 struct list_head list;
372 struct zone_reclaim_stat reclaim_stat;
374 unsigned long pages_scanned; /* since last reclaim */
375 unsigned long flags; /* zone flags, see below */
377 /* Zone statistics */
378 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
381 * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
382 * this zone's LRU. Maintained by the pageout code.
384 unsigned int inactive_ratio;
388 /* Rarely used or read-mostly fields */
391 * wait_table -- the array holding the hash table
392 * wait_table_hash_nr_entries -- the size of the hash table array
393 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
395 * The purpose of all these is to keep track of the people
396 * waiting for a page to become available and make them
397 * runnable again when possible. The trouble is that this
398 * consumes a lot of space, especially when so few things
399 * wait on pages at a given time. So instead of using
400 * per-page waitqueues, we use a waitqueue hash table.
402 * The bucket discipline is to sleep on the same queue when
403 * colliding and wake all in that wait queue when removing.
404 * When something wakes, it must check to be sure its page is
405 * truly available, a la thundering herd. The cost of a
406 * collision is great, but given the expected load of the
407 * table, they should be so rare as to be outweighed by the
408 * benefits from the saved space.
410 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
411 * primary users of these fields, and in mm/page_alloc.c
412 * free_area_init_core() performs the initialization of them.
414 wait_queue_head_t * wait_table;
415 unsigned long wait_table_hash_nr_entries;
416 unsigned long wait_table_bits;
419 * Discontig memory support fields.
421 struct pglist_data *zone_pgdat;
422 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
423 unsigned long zone_start_pfn;
426 * zone_start_pfn, spanned_pages and present_pages are all
427 * protected by span_seqlock. It is a seqlock because it has
428 * to be read outside of zone->lock, and it is done in the main
429 * allocator path. But, it is written quite infrequently.
431 * The lock is declared along with zone->lock because it is
432 * frequently read in proximity to zone->lock. It's good to
433 * give them a chance of being in the same cacheline.
435 unsigned long spanned_pages; /* total size, including holes */
436 unsigned long present_pages; /* amount of memory (excluding holes) */
439 * rarely used fields:
442 } ____cacheline_internodealigned_in_smp;
445 ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */
446 ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */
447 ZONE_CONGESTED, /* zone has many dirty pages backed by
452 static inline void zone_set_flag(struct zone *zone, zone_flags_t flag)
454 set_bit(flag, &zone->flags);
457 static inline int zone_test_and_set_flag(struct zone *zone, zone_flags_t flag)
459 return test_and_set_bit(flag, &zone->flags);
462 static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag)
464 clear_bit(flag, &zone->flags);
467 static inline int zone_is_reclaim_congested(const struct zone *zone)
469 return test_bit(ZONE_CONGESTED, &zone->flags);
472 static inline int zone_is_reclaim_locked(const struct zone *zone)
474 return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
477 static inline int zone_is_oom_locked(const struct zone *zone)
479 return test_bit(ZONE_OOM_LOCKED, &zone->flags);
483 * The "priority" of VM scanning is how much of the queues we will scan in one
484 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
485 * queues ("queue_length >> 12") during an aging round.
487 #define DEF_PRIORITY 12
489 /* Maximum number of zones on a zonelist */
490 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
495 * The NUMA zonelists are doubled because we need zonelists that restrict the
496 * allocations to a single node for GFP_THISNODE.
498 * [0] : Zonelist with fallback
499 * [1] : No fallback (GFP_THISNODE)
501 #define MAX_ZONELISTS 2
505 * We cache key information from each zonelist for smaller cache
506 * footprint when scanning for free pages in get_page_from_freelist().
508 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
509 * up short of free memory since the last time (last_fullzone_zap)
510 * we zero'd fullzones.
511 * 2) The array z_to_n[] maps each zone in the zonelist to its node
512 * id, so that we can efficiently evaluate whether that node is
513 * set in the current tasks mems_allowed.
515 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
516 * indexed by a zones offset in the zonelist zones[] array.
518 * The get_page_from_freelist() routine does two scans. During the
519 * first scan, we skip zones whose corresponding bit in 'fullzones'
520 * is set or whose corresponding node in current->mems_allowed (which
521 * comes from cpusets) is not set. During the second scan, we bypass
522 * this zonelist_cache, to ensure we look methodically at each zone.
524 * Once per second, we zero out (zap) fullzones, forcing us to
525 * reconsider nodes that might have regained more free memory.
526 * The field last_full_zap is the time we last zapped fullzones.
528 * This mechanism reduces the amount of time we waste repeatedly
529 * reexaming zones for free memory when they just came up low on
530 * memory momentarilly ago.
532 * The zonelist_cache struct members logically belong in struct
533 * zonelist. However, the mempolicy zonelists constructed for
534 * MPOL_BIND are intentionally variable length (and usually much
535 * shorter). A general purpose mechanism for handling structs with
536 * multiple variable length members is more mechanism than we want
537 * here. We resort to some special case hackery instead.
539 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
540 * part because they are shorter), so we put the fixed length stuff
541 * at the front of the zonelist struct, ending in a variable length
542 * zones[], as is needed by MPOL_BIND.
544 * Then we put the optional zonelist cache on the end of the zonelist
545 * struct. This optional stuff is found by a 'zlcache_ptr' pointer in
546 * the fixed length portion at the front of the struct. This pointer
547 * both enables us to find the zonelist cache, and in the case of
548 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
549 * to know that the zonelist cache is not there.
551 * The end result is that struct zonelists come in two flavors:
552 * 1) The full, fixed length version, shown below, and
553 * 2) The custom zonelists for MPOL_BIND.
554 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
556 * Even though there may be multiple CPU cores on a node modifying
557 * fullzones or last_full_zap in the same zonelist_cache at the same
558 * time, we don't lock it. This is just hint data - if it is wrong now
559 * and then, the allocator will still function, perhaps a bit slower.
563 struct zonelist_cache {
564 unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */
565 DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */
566 unsigned long last_full_zap; /* when last zap'd (jiffies) */
569 #define MAX_ZONELISTS 1
570 struct zonelist_cache;
574 * This struct contains information about a zone in a zonelist. It is stored
575 * here to avoid dereferences into large structures and lookups of tables
578 struct zone *zone; /* Pointer to actual zone */
579 int zone_idx; /* zone_idx(zoneref->zone) */
583 * One allocation request operates on a zonelist. A zonelist
584 * is a list of zones, the first one is the 'goal' of the
585 * allocation, the other zones are fallback zones, in decreasing
588 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
589 * as explained above. If zlcache_ptr is NULL, there is no zlcache.
591 * To speed the reading of the zonelist, the zonerefs contain the zone index
592 * of the entry being read. Helper functions to access information given
593 * a struct zoneref are
595 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
596 * zonelist_zone_idx() - Return the index of the zone for an entry
597 * zonelist_node_idx() - Return the index of the node for an entry
600 struct zonelist_cache *zlcache_ptr; // NULL or &zlcache
601 struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
603 struct zonelist_cache zlcache; // optional ...
607 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
608 struct node_active_region {
609 unsigned long start_pfn;
610 unsigned long end_pfn;
613 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
615 #ifndef CONFIG_DISCONTIGMEM
616 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
617 extern struct page *mem_map;
621 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
622 * (mostly NUMA machines?) to denote a higher-level memory zone than the
625 * On NUMA machines, each NUMA node would have a pg_data_t to describe
626 * it's memory layout.
628 * Memory statistics and page replacement data structures are maintained on a
632 typedef struct pglist_data {
633 struct zone node_zones[MAX_NR_ZONES];
634 struct zonelist node_zonelists[MAX_ZONELISTS];
636 #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
637 struct page *node_mem_map;
638 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
639 struct page_cgroup *node_page_cgroup;
642 #ifndef CONFIG_NO_BOOTMEM
643 struct bootmem_data *bdata;
645 #ifdef CONFIG_MEMORY_HOTPLUG
647 * Must be held any time you expect node_start_pfn, node_present_pages
648 * or node_spanned_pages stay constant. Holding this will also
649 * guarantee that any pfn_valid() stays that way.
651 * Nests above zone->lock and zone->size_seqlock.
653 spinlock_t node_size_lock;
655 unsigned long node_start_pfn;
656 unsigned long node_present_pages; /* total number of physical pages */
657 unsigned long node_spanned_pages; /* total size of physical page
658 range, including holes */
660 wait_queue_head_t kswapd_wait;
661 struct task_struct *kswapd;
662 int kswapd_max_order;
663 enum zone_type classzone_idx;
666 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
667 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
668 #ifdef CONFIG_FLAT_NODE_MEM_MAP
669 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
671 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
673 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
675 #define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn)
677 #define node_end_pfn(nid) ({\
678 pg_data_t *__pgdat = NODE_DATA(nid);\
679 __pgdat->node_start_pfn + __pgdat->node_spanned_pages;\
682 #include <linux/memory_hotplug.h>
684 extern struct mutex zonelists_mutex;
685 void build_all_zonelists(void *data);
686 void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx);
687 bool zone_watermark_ok(struct zone *z, int order, unsigned long mark,
688 int classzone_idx, int alloc_flags);
689 bool zone_watermark_ok_safe(struct zone *z, int order, unsigned long mark,
690 int classzone_idx, int alloc_flags);
691 enum memmap_context {
695 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
697 enum memmap_context context);
699 #ifdef CONFIG_HAVE_MEMORY_PRESENT
700 void memory_present(int nid, unsigned long start, unsigned long end);
702 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
705 #ifdef CONFIG_HAVE_MEMORYLESS_NODES
706 int local_memory_node(int node_id);
708 static inline int local_memory_node(int node_id) { return node_id; };
711 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
712 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
716 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
718 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
720 static inline int populated_zone(struct zone *zone)
722 return (!!zone->present_pages);
725 extern int movable_zone;
727 static inline int zone_movable_is_highmem(void)
729 #if defined(CONFIG_HIGHMEM) && defined(CONFIG_HAVE_MEMBLOCK_NODE)
730 return movable_zone == ZONE_HIGHMEM;
736 static inline int is_highmem_idx(enum zone_type idx)
738 #ifdef CONFIG_HIGHMEM
739 return (idx == ZONE_HIGHMEM ||
740 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
746 static inline int is_normal_idx(enum zone_type idx)
748 return (idx == ZONE_NORMAL);
752 * is_highmem - helper function to quickly check if a struct zone is a
753 * highmem zone or not. This is an attempt to keep references
754 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
755 * @zone - pointer to struct zone variable
757 static inline int is_highmem(struct zone *zone)
759 #ifdef CONFIG_HIGHMEM
760 int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones;
761 return zone_off == ZONE_HIGHMEM * sizeof(*zone) ||
762 (zone_off == ZONE_MOVABLE * sizeof(*zone) &&
763 zone_movable_is_highmem());
769 static inline int is_normal(struct zone *zone)
771 return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
774 static inline int is_dma32(struct zone *zone)
776 #ifdef CONFIG_ZONE_DMA32
777 return zone == zone->zone_pgdat->node_zones + ZONE_DMA32;
783 static inline int is_dma(struct zone *zone)
785 #ifdef CONFIG_ZONE_DMA
786 return zone == zone->zone_pgdat->node_zones + ZONE_DMA;
792 /* These two functions are used to setup the per zone pages min values */
794 int min_free_kbytes_sysctl_handler(struct ctl_table *, int,
795 void __user *, size_t *, loff_t *);
796 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
797 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int,
798 void __user *, size_t *, loff_t *);
799 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
800 void __user *, size_t *, loff_t *);
801 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
802 void __user *, size_t *, loff_t *);
803 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
804 void __user *, size_t *, loff_t *);
806 extern int numa_zonelist_order_handler(struct ctl_table *, int,
807 void __user *, size_t *, loff_t *);
808 extern char numa_zonelist_order[];
809 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
811 #ifndef CONFIG_NEED_MULTIPLE_NODES
813 extern struct pglist_data contig_page_data;
814 #define NODE_DATA(nid) (&contig_page_data)
815 #define NODE_MEM_MAP(nid) mem_map
817 #else /* CONFIG_NEED_MULTIPLE_NODES */
819 #include <asm/mmzone.h>
821 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
823 extern struct pglist_data *first_online_pgdat(void);
824 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
825 extern struct zone *next_zone(struct zone *zone);
828 * for_each_online_pgdat - helper macro to iterate over all online nodes
829 * @pgdat - pointer to a pg_data_t variable
831 #define for_each_online_pgdat(pgdat) \
832 for (pgdat = first_online_pgdat(); \
834 pgdat = next_online_pgdat(pgdat))
836 * for_each_zone - helper macro to iterate over all memory zones
837 * @zone - pointer to struct zone variable
839 * The user only needs to declare the zone variable, for_each_zone
842 #define for_each_zone(zone) \
843 for (zone = (first_online_pgdat())->node_zones; \
845 zone = next_zone(zone))
847 #define for_each_populated_zone(zone) \
848 for (zone = (first_online_pgdat())->node_zones; \
850 zone = next_zone(zone)) \
851 if (!populated_zone(zone)) \
855 static inline struct zone *zonelist_zone(struct zoneref *zoneref)
857 return zoneref->zone;
860 static inline int zonelist_zone_idx(struct zoneref *zoneref)
862 return zoneref->zone_idx;
865 static inline int zonelist_node_idx(struct zoneref *zoneref)
868 /* zone_to_nid not available in this context */
869 return zoneref->zone->node;
872 #endif /* CONFIG_NUMA */
876 * next_zones_zonelist - Returns the next zone at or below highest_zoneidx within the allowed nodemask using a cursor within a zonelist as a starting point
877 * @z - The cursor used as a starting point for the search
878 * @highest_zoneidx - The zone index of the highest zone to return
879 * @nodes - An optional nodemask to filter the zonelist with
880 * @zone - The first suitable zone found is returned via this parameter
882 * This function returns the next zone at or below a given zone index that is
883 * within the allowed nodemask using a cursor as the starting point for the
884 * search. The zoneref returned is a cursor that represents the current zone
885 * being examined. It should be advanced by one before calling
886 * next_zones_zonelist again.
888 struct zoneref *next_zones_zonelist(struct zoneref *z,
889 enum zone_type highest_zoneidx,
894 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
895 * @zonelist - The zonelist to search for a suitable zone
896 * @highest_zoneidx - The zone index of the highest zone to return
897 * @nodes - An optional nodemask to filter the zonelist with
898 * @zone - The first suitable zone found is returned via this parameter
900 * This function returns the first zone at or below a given zone index that is
901 * within the allowed nodemask. The zoneref returned is a cursor that can be
902 * used to iterate the zonelist with next_zones_zonelist by advancing it by
903 * one before calling.
905 static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
906 enum zone_type highest_zoneidx,
910 return next_zones_zonelist(zonelist->_zonerefs, highest_zoneidx, nodes,
915 * for_each_zone_zonelist_nodemask - helper macro to iterate over valid zones in a zonelist at or below a given zone index and within a nodemask
916 * @zone - The current zone in the iterator
917 * @z - The current pointer within zonelist->zones being iterated
918 * @zlist - The zonelist being iterated
919 * @highidx - The zone index of the highest zone to return
920 * @nodemask - Nodemask allowed by the allocator
922 * This iterator iterates though all zones at or below a given zone index and
923 * within a given nodemask
925 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
926 for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
928 z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \
931 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
932 * @zone - The current zone in the iterator
933 * @z - The current pointer within zonelist->zones being iterated
934 * @zlist - The zonelist being iterated
935 * @highidx - The zone index of the highest zone to return
937 * This iterator iterates though all zones at or below a given zone index.
939 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
940 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
942 #ifdef CONFIG_SPARSEMEM
943 #include <asm/sparsemem.h>
946 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
947 !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
948 static inline unsigned long early_pfn_to_nid(unsigned long pfn)
954 #ifdef CONFIG_FLATMEM
955 #define pfn_to_nid(pfn) (0)
958 #ifdef CONFIG_SPARSEMEM
961 * SECTION_SHIFT #bits space required to store a section #
963 * PA_SECTION_SHIFT physical address to/from section number
964 * PFN_SECTION_SHIFT pfn to/from section number
966 #define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
968 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
969 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
971 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
973 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
974 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
976 #define SECTION_BLOCKFLAGS_BITS \
977 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
979 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
980 #error Allocator MAX_ORDER exceeds SECTION_SIZE
983 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
984 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
986 #define SECTION_ALIGN_UP(pfn) (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
987 #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
993 * This is, logically, a pointer to an array of struct
994 * pages. However, it is stored with some other magic.
995 * (see sparse.c::sparse_init_one_section())
997 * Additionally during early boot we encode node id of
998 * the location of the section here to guide allocation.
999 * (see sparse.c::memory_present())
1001 * Making it a UL at least makes someone do a cast
1002 * before using it wrong.
1004 unsigned long section_mem_map;
1006 /* See declaration of similar field in struct zone */
1007 unsigned long *pageblock_flags;
1008 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
1010 * If !SPARSEMEM, pgdat doesn't have page_cgroup pointer. We use
1011 * section. (see memcontrol.h/page_cgroup.h about this.)
1013 struct page_cgroup *page_cgroup;
1018 #ifdef CONFIG_SPARSEMEM_EXTREME
1019 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
1021 #define SECTIONS_PER_ROOT 1
1024 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
1025 #define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
1026 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
1028 #ifdef CONFIG_SPARSEMEM_EXTREME
1029 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
1031 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
1034 static inline struct mem_section *__nr_to_section(unsigned long nr)
1036 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
1038 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
1040 extern int __section_nr(struct mem_section* ms);
1041 extern unsigned long usemap_size(void);
1044 * We use the lower bits of the mem_map pointer to store
1045 * a little bit of information. There should be at least
1046 * 3 bits here due to 32-bit alignment.
1048 #define SECTION_MARKED_PRESENT (1UL<<0)
1049 #define SECTION_HAS_MEM_MAP (1UL<<1)
1050 #define SECTION_MAP_LAST_BIT (1UL<<2)
1051 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
1052 #define SECTION_NID_SHIFT 2
1054 static inline struct page *__section_mem_map_addr(struct mem_section *section)
1056 unsigned long map = section->section_mem_map;
1057 map &= SECTION_MAP_MASK;
1058 return (struct page *)map;
1061 static inline int present_section(struct mem_section *section)
1063 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
1066 static inline int present_section_nr(unsigned long nr)
1068 return present_section(__nr_to_section(nr));
1071 static inline int valid_section(struct mem_section *section)
1073 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
1076 static inline int valid_section_nr(unsigned long nr)
1078 return valid_section(__nr_to_section(nr));
1081 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1083 return __nr_to_section(pfn_to_section_nr(pfn));
1086 #ifndef CONFIG_HAVE_ARCH_PFN_VALID
1087 static inline int pfn_valid(unsigned long pfn)
1089 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1091 return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
1095 static inline int pfn_present(unsigned long pfn)
1097 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1099 return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1103 * These are _only_ used during initialisation, therefore they
1104 * can use __initdata ... They could have names to indicate
1108 #define pfn_to_nid(pfn) \
1110 unsigned long __pfn_to_nid_pfn = (pfn); \
1111 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1114 #define pfn_to_nid(pfn) (0)
1117 #define early_pfn_valid(pfn) pfn_valid(pfn)
1118 void sparse_init(void);
1120 #define sparse_init() do {} while (0)
1121 #define sparse_index_init(_sec, _nid) do {} while (0)
1122 #endif /* CONFIG_SPARSEMEM */
1124 #ifdef CONFIG_NODES_SPAN_OTHER_NODES
1125 bool early_pfn_in_nid(unsigned long pfn, int nid);
1127 #define early_pfn_in_nid(pfn, nid) (1)
1130 #ifndef early_pfn_valid
1131 #define early_pfn_valid(pfn) (1)
1134 void memory_present(int nid, unsigned long start, unsigned long end);
1135 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1138 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1139 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1140 * pfn_valid_within() should be used in this case; we optimise this away
1141 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1143 #ifdef CONFIG_HOLES_IN_ZONE
1144 #define pfn_valid_within(pfn) pfn_valid(pfn)
1146 #define pfn_valid_within(pfn) (1)
1149 #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1151 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1152 * associated with it or not. In FLATMEM, it is expected that holes always
1153 * have valid memmap as long as there is valid PFNs either side of the hole.
1154 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1157 * However, an ARM, and maybe other embedded architectures in the future
1158 * free memmap backing holes to save memory on the assumption the memmap is
1159 * never used. The page_zone linkages are then broken even though pfn_valid()
1160 * returns true. A walker of the full memmap must then do this additional
1161 * check to ensure the memmap they are looking at is sane by making sure
1162 * the zone and PFN linkages are still valid. This is expensive, but walkers
1163 * of the full memmap are extremely rare.
1165 int memmap_valid_within(unsigned long pfn,
1166 struct page *page, struct zone *zone);
1168 static inline int memmap_valid_within(unsigned long pfn,
1169 struct page *page, struct zone *zone)
1173 #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1175 #endif /* !__GENERATING_BOUNDS.H */
1176 #endif /* !__ASSEMBLY__ */
1177 #endif /* _LINUX_MMZONE_H */